Study On Cell Variations Of Lithium Ion Power Battery Packs In Electric Vehicles

A battery pack system in an electric vehicle usually consists of hundreds of singlecells connected in series and parallel. Due to the inconsistent manufacturing process andthe inhomogeneous operating environment, cells always have variations which cannotbe eliminated. Significant degradation in energy density, cycle life and safety isobserved after pack construction due to cell variations.This dissertation focuses on power battery packs which bridge electric vehicles andsingle cells. Cell variation is the key issue between a single cell and a battery pack. Inview of shorter cycle life of battery packs, the mechanism and impact factors of packlife evolution are studied. Cell variation identification methods, as well as thecorresponding equalization algorithms and fault diagnosis methods for electric vehiclesduring constant current charge and dynamic discharge period are respectively studied.Firstly, a category of cell aging mechanism is reviewed and two-dimensional cellaging mechanism based on electric quantity scatter diagram is proposed. The disciplineof pack capacity evolution is consequently revealed. Battery pack experiment andsimulation are used to verify the discipline of pack capacity evolution. The disciplinesuggests that packs age faster than single cells because the pack capacity loss is the sumof the cell capacity loss who has the minimum remaining charge electric quantity andthe loss of lithium inventory difference at anode between cells. It is also revealed bysimulation that pack capacity fades increasingly. Besides, a series of single parametersimulation studies indicate that coulombic efficiency and temperature are the dominatefactors which impact cell variation.Secondly, uniform charging cell voltage curve hypothesis is raised for constantcharging process. By charging cell voltage curve transformation and transformationparameter optimization, cell SOCs (State of Charge) and capacities are identified, andpack capacity is estimated subsequently. By analyzing on the electric quantity scatterdiagram, on line equalization algorithms by conditioning charging cell voltage curvesbased on electric quantity are further proposed. By using uniform charging cell voltagecurve hypothesis, the equalization algorithm based on remaining charging electricquantity estimation and an adaptive fuzzy logic equalization algorithm which fuzzilyjudges cell capacities and SOCs by voltages at the start and end of charging arerespectively proposed. The simulation results show that the proposed equalization algorithms are feasible and over equalization can be avoided.Finally, a frequency division battery pack model is established for dynamicdischarge period: the high frequency cell mean model is used to study overallperformance of the battery pack; the low frequency cell difference model is oriented tothe study of cell variation. On one hand, SOC variation can be identified in hybridelectric vehicles by intermittently lowering SOC maintaining target to the identificationzone with cell difference model identification of OCV (Open Circuit Voltage) difference.On the other hand, dynamic discharge periods of battery pack history data are analyzedby cell difference model identification of total resistance difference. Along withShannon entropy, internal resistance fault or contact resistance fault is distinguished.